Orchestration

Ra conducts the orchestra. Fifteen models. One pipeline. One bill. One brief.

Orchestration turns a brief into a finished video. Ra picks the best model for each primitive and routes them as a single, coherent production.

The model stack Ra orchestrates

Primitive	Models Ra routes between
Keyframes / image generation	Seedance, Midjourney, Flux, Ideogram
Motion / video generation	Veo 3, Kling 2.5, Runway, Pika, Luma, Hailuo
Dialogue / voice synthesis	ElevenLabs, Cartesia, OpenAI TTS
Audio score and SFX	Suno, Udio, generative SFX layers
Captions and dubbing	Whisper-class transcription plus multilingual TTS
Post and effects	Native effects pipeline, frame interpolation, upscaling

One bill covers the entire stack

You never subscribe to Runway and Pika and ElevenLabs and Kling separately. You never pick a model. You never guess which is best this week. Ra knows, because Ra is routing thousands of shots per day across all of them and learning which model wins for which kind of shot.

What orchestration does for you

Ra handles	So you don’t have to
Picks the best model per shot by style, subject, motion type, outcome goal	Research fifteen models every time you need motion
Maintains character consistency across model boundaries	Re-upload character sheets every scene
Routes around model outages and rate limits	Wake up to a failed render at 3 AM
Upgrades to new models transparently (Veo 4 ships → Ra uses it)	Migrate your workflow when a new model lands
Negotiates enterprise rates on your behalf	Pay retail for every model subscription
Unifies output — same color grade, same voice, same pacing	Stitch five different-looking outputs in an editor afterward

Can you override?

Yes. Per scene, per project, or globally. See How Ra picks a model for the routing logic and override options.

Related

• Pillars overview
• Intelligence
• Memory
• How Ra picks a model
• Credits